Revisiting reference intervals in transfusion medicine: Innovations and challenges in current practice

Reference intervals (RIs) are essential for donor selection, blood component quality, and post-transfusion safety in transfusion medicine. The CLSI EP28 non-parametric method, based on the Harris-Boyd framework, is widely used but has limitations in handling outliers and skewed distributions. Emerging computational approaches, including Expectation-Maximization (EM), reflimR, and refineR, offer improved RI estimation. This study evaluates the limitations of CLSI EP28 in transfusion medicine and compares EM, reflimR, and refineR to assess their effectiveness in refining RI estimation.A simulated blood donor dataset (n = 500) was generated, modelling hemoglobin (Hb), hematocrit (Hct), and platelet count (PLT). Four RI estimation methods were compared: (1) CLSI EP28 (Non-Parametric Percentile Method), (2) Expectation-Maximization (EM) Algorithm, (3) reflimR (Robust Regression Method) and, (4) refineR (Advanced Statistical Modeling Method).The dataset had mean values of 15 g/dL (Hb), 44 % (Hct), and 250 × 10⁹/L (PLT). CLSI EP28 and EM yielded similar RIs, relying on empirical percentiles. ReflimR and refineR produced wider RIs, improving outlier resistance and distributional accuracy.While CLSI EP28 remains the regulatory standard, computational RI estimation methods improve accuracy, robustness, and regulatory compliance (IVDR, FDA 510(k), ISO 15189). Implementing EM, reflimR, and refineR can enhance donor screening and transfusion safety.